Constrained model predictive control for enhanced trajectory tracking in multi-DOF robotic manipulators

Abstract

Controlling a multi-degree-of-freedom (multi-DOF) robotic manipulator is complicated by nonlinear dynamics, coupled joints, and constraints such as joint limits, actuator saturation, and collision avoidance. The focus of this proposed work is the development and implementation of constrained model predictive control (MPC) algorithms for robotic manipulators. The key features of this proposal include the use of the dynamics of the manipulator in the process of prediction and the ability for the controller to take optimal actions over a fixed time horizon, while ensuring that the full range of physical and safety constraints is satisfied. The proposed MPC framework incorporates a discrete-time state-space model of the robotic manipulator that can be optimized using quadratic programming (QP), which allows for the model to be expressed in a general stable form to enable optimization. Linear and nonlinear MPC approaches will be considered, but the emphasis will be on the feasibility of real-time implementation and robustness of the controller to modelling errors and disturbances from the environment. The algorithm can be used in simulation and on a physical multi-DOF robotic arm in applications ranging from trajectory tracking to obstacle avoidance and precision positioning of the end-effector. Compared to traditional control techniques like PID, and computed torque control proves the superiority of MPC in controlling dynamic constraints and increasing control accuracy. The research also discusses implementation techniques involving reduced-order models and efficient solvers to address real-time computational needs, enabling safe and effective deployment in sophisticated robotic devices.